This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. The ClinicalTrials.gov record includes a posted primary outcome measure but no formal statistical-analysis records, so no treatment-effect estimate, confidence interval, or p-value is attributed to the registry.
1. Trial at a Glance
BIRCH was a completed phase 2, single-group study of atezolizumab in participants with PD-L1-positive locally advanced or metastatic non-small cell lung cancer. The registry reports an enrollment of 667 participants across three treatment-line cohorts and lists objective response rate as the registered primary endpoint.
| Feature | BIRCH |
|---|---|
| Trial name | BIRCH |
| NCT identifier | NCT02031458 |
| Phase | Phase 2 |
| Status | Completed |
| Condition | Non-Small Cell Lung Cancer |
| Intervention | Atezolizumab |
| Allocation | NA |
| Design model | Single-group |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 667 |
| Arms | 1 |
| Lead sponsor | Hoffmann-La Roche |
| Sponsor type | Industry |
| Trial start | 2014-01-22 |
| Primary completion | 2015-05-28 |
2. Clinical Question
The BIRCH trial addressed the use of atezolizumab in participants with PD-L1-positive locally advanced or metastatic non-small cell lung cancer. The registered primary question was operationalized through the percentage of participants achieving an objective response according to RECIST v1.1 as assessed by an independent review facility.
Population
Participants with programmed death-ligand 1 (PD-L1)-positive locally advanced or metastatic non-small cell lung cancer.
Intervention
Atezolizumab.
Comparator
The study is registered as single-group, with one arm and no allocation to a separate comparator arm.
Primary question
What percentage of participants achieve an objective response by the registered RECIST v1.1 and independent-review-facility assessment framework?
3. Trial Design
Atezolizumab
- Intervention: atezolizumab
- One study arm
- No separate comparator arm is reported in the registry design data
Treatment-line cohorts
- Cohort 1: First Line Atezolizumab
- Cohort 2: Second Line Atezolizumab
- Cohort 3: Third Line and Beyond Atezoliz
Enrollment and study timing
| Registry feature | Reported value |
|---|---|
| Start | 2014-01-22 |
| Primary completion | 2015-05-28 |
| Enrollment | 667 |
| Study status | Completed |
4. Treatment-Line Cohorts and Safety Population Information
The ClinicalTrials.gov record identifies three cohorts and provide serious-adverse-event counts by cohort. These counts are presented as affected participants over participants at risk. Because the study is single-group, these are cohort-specific safety summaries rather than randomized comparisons between treatment groups.
| Cohort | Serious adverse events affected / at risk | Descriptive interpretation |
|---|---|---|
| Cohort 1: First Line Atezolizumab | 47/138 | 47 affected participants among 138 at risk |
| Cohort 2: Second Line Atezolizumab | 117/269 | 117 affected participants among 269 at risk |
| Cohort 3: Third Line and Beyond Atezoliz | 113/252 | 113 affected participants among 252 at risk |
5. Registered Primary Endpoint
| Endpoint | Registry definition / time frame | Endpoint type |
|---|---|---|
| Percentage of Participants Achieving Objective Response (ORR) Per Response Evaluation Criteria In Solid Tumors (RECIST) Version (v) 1.1 as Assessed by Independent Review Facility (IRF) | Screening, Every 6 weeks (± 3 days) for 12 months following Cycle 1, Day 1 and every 9 weeks (± 1 week) thereafter until disease progression, intolerable toxicity or death until data cut-off on 28 May 2015 (Up to 16 months) | Binary |
Registry definition
The registry definition states that ORR was the percentage of participants whose confirmed best overall response was either a Partial Response (PR) or a Complete Response (CR) based upon the IRF assessment per RECIST v1.1.
The registry-reported definition further specifies that a complete response involved disappearance of all target and non-target lesions, with pathological lymph nodes having reduction in short axis to less than 10 millimeters. The registry text then begins the partial-response definition but ends after the phrase “greater than or equal to,” so no additional PR threshold is added here.
This is a binary response endpoint: each participant is classified according to whether the registered objective-response criterion was achieved. The percentage is therefore a proportion rather than a time-to-event hazard ratio or a continuous change score.
6. Reported Results and Planned Statistical Analysis
ClinicalTrials.gov reports that results are posted for the primary endpoint, but the ClinicalTrials.gov record contains no formal statistical analyses. Consequently, there is no registry-provided formal comparison, effect estimate, confidence interval, or p-value that can be reproduced on this page.
Planned Analysis
For a single-group binary endpoint such as ORR, the natural descriptive analysis is the observed proportion of participants achieving objective response:
The numerator represents participants meeting the registered response definition. The denominator should be defined according to the prespecified analysis population and response-evaluation rules.
A confidence interval for a single proportion would normally accompany the point estimate. Depending on the statistical analysis plan, an exact binomial interval or another prespecified binomial interval could be used. A formal one-sample hypothesis test could also compare the observed response proportion with a prespecified benchmark if such a benchmark had been specified.
None of those additional numerical quantities is reported in the ClinicalTrials.gov record. They should therefore be treated here as standard analytical approaches for this endpoint type, not as the actual BIRCH statistical analysis.
Why the single-group design changes the question
Because BIRCH is registered as a single-group study with one arm, its primary endpoint does not naturally produce a randomized treatment-versus-control contrast. The central statistical object is a response proportion, potentially compared with a prespecified reference value if the protocol defines one.
A single-group ORR describes the observed frequency of objective response under the study's eligibility, assessment, and follow-up framework. It does not, by itself, estimate how much better atezolizumab performs than an alternative treatment. Without a randomized comparator or a clearly specified external benchmark, the observed proportion should not be converted into a comparative treatment effect.
7. Statistical Methodology
Binary endpoint analysis
The primary endpoint is explicitly identified as binary. Each participant's outcome is conceptually reduced to whether the registered objective-response criterion was achieved or not achieved. The resulting data can be represented by a binomial count: a number of responders among a defined number of evaluable participants.
Here, x is the number of participants classified as responders and n is the relevant analysis denominator. The ClinicalTrials.gov record does not provide these numerical values.
Confidence intervals for a proportion
A point estimate alone does not describe statistical precision. A confidence interval quantifies uncertainty around the estimated response proportion under the chosen binomial model. For a clinical-trial response endpoint, an exact binomial interval is one commonly used approach, particularly when the number of responses is small or when normal approximations may be inadequate.
Independent review assessment
The registered endpoint specifies assessment by an Independent Review Facility (IRF). This matters statistically because the endpoint classification is based on an independent assessment process rather than being described simply as investigator-assessed response in the registry definition.
Repeated response assessments
The registered time frame specifies assessments beginning at screening and occurring every 6 weeks with a tolerance of ±3 days for 12 months following Cycle 1, Day 1, followed by assessments every 9 weeks with a tolerance of ±1 week thereafter until the registry text ends. Repeated radiologic assessments create the longitudinal information from which a participant's best overall response can be determined.
Confirmed response
The registry defines ORR using a confirmed best overall response. Statistically, this is important because response classification is not simply a single measurement at one arbitrary visit. The participant must satisfy the registered response framework for the response to enter the primary endpoint.
What the analysis estimates
The proportion of participants satisfying the registered objective-response definition.
What it does not estimate
A randomized relative treatment effect, because the registry-reported design contains no separate comparator arm.
What determines precision
The number of evaluable participants and the number classified as responders.
What determines validity
The prespecified response definition, assessment schedule, analysis population, and handling of incomplete response assessments.
8. Statistical Methods Explained
Why is ORR treated as a binary endpoint?
The registry classifies the primary endpoint as binary because each participant ultimately contributes a response classification: the participant either has a confirmed complete or partial response under the registered framework or does not. This permits analysis using binomial methods rather than continuous-outcome or time-to-event methods.
What does an ORR percentage mean?
An ORR percentage is a proportion. For example, conceptually, an ORR of 30% would mean that 30% of the relevant analysis population met the registered response criterion. It would not mean that every participant experienced a 30% tumor reduction, nor would it describe duration of response.
Why is the denominator important?
The percentage depends on who is included in the denominator. In a trial with repeated assessments and potentially incomplete evaluations, the distinction between all enrolled participants and the prespecified response-evaluable population can materially affect the reported percentage. The ClinicalTrials.gov record provides an enrollment of 667 but do not provide the numerical response-analysis denominator.
Why use a confidence interval?
The observed response proportion is calculated from a finite number of participants. A confidence interval communicates uncertainty around that estimate. A narrow interval generally indicates greater statistical precision than a wide interval, although the interval does not describe the distribution of responses in future individual patients.
Why is an independent review facility relevant?
Because response classification can be influenced by radiologic interpretation, the registry's specification of an independent review facility defines an important part of the endpoint-assessment process. The statistical analysis is only as reproducible as the clinical rules used to classify participants as responders.
Why can the primary endpoint not be summarized with a hazard ratio?
Hazard ratios are designed for time-to-event comparisons. The registered primary endpoint here is a binary objective-response measure. A hazard ratio would therefore answer a different question unless a separate time-to-response or survival endpoint were being analyzed.
9. Response Assessment as a Longitudinal Process
Although the primary endpoint is binary, the underlying assessment process is longitudinal. Participants are assessed repeatedly according to the registered schedule, and the endpoint uses a confirmed best overall response. This creates an important distinction between the measurement process and the final statistical endpoint.
Baseline assessment
The registered primary endpoint begins its assessment framework at screening.
First 12 months
Response assessments are scheduled every 6 weeks with the registry-specified timing tolerance following Cycle 1, Day 1.
Thereafter
The registry specifies a less frequent assessment schedule after the first 12 months.
Confirmed best overall response
The final binary endpoint is based on whether the participant achieved a confirmed CR or PR according to the registered IRF/RECIST v1.1 framework.
10. Safety Results
The ClinicalTrials.gov record reports serious adverse events by the three treatment-line cohorts. These are the safety results available in the ClinicalTrials.gov record and are presented without constructing an unreported comparative analysis.
| Cohort | Affected | At risk | Registry description |
|---|---|---|---|
| Cohort 1: First Line Atezolizumab | 47 | 138 | 47/138 participants affected |
| Cohort 2: Second Line Atezolizumab | 117 | 269 | 117/269 participants affected |
| Cohort 3: Third Line and Beyond Atezoliz | 113 | 252 | 113/252 participants affected |
How to interpret these safety counts
The numerator identifies the number of participants affected by serious adverse events and the denominator identifies the participants at risk for that cohort. These counts are descriptive. Because all three cohorts received the same registered intervention and the study is single-group, the ClinicalTrials.gov record does not support a randomized relative-risk or odds-ratio comparison.
A serious-adverse-event count is a safety frequency, not an estimate of treatment efficacy. It should be interpreted in relation to the relevant at-risk population and the trial's exposure and follow-up structure. The ClinicalTrials.gov record does not provide enough information to determine whether the cohorts differed statistically in serious-adverse-event risk.
11. What the Single-Group Design Does — and Does Not — Mean
It can estimate
The observed response proportion within the enrolled study population, subject to the prespecified response-evaluation framework.
It cannot directly estimate
The causal difference in ORR between atezolizumab and a concurrently randomized comparator arm.
External comparison
A comparison with historical or external data requires an explicitly defined reference population and assumptions that are not reported here.
Internal validity
Without randomization between treatment groups, the design does not provide the same protection against comparative confounding as a randomized controlled trial.
This distinction is central to interpreting a phase 2 single-group study. A high response proportion can be scientifically informative, but the strength of any comparative claim depends on the source of the comparator and the assumptions needed to make the comparison.
12. Limitations
- Single-group design: the registry design contains one arm and no randomized comparator, limiting direct causal comparisons with alternative treatments.
- No formal statistical-analysis records reported: ClinicalTrials.gov indicates that results are posted, but the ClinicalTrials.gov record contains no statistical-analysis entries for the primary endpoint.
- Incomplete numerical endpoint results in the ClinicalTrials.gov record: the ClinicalTrials.gov recordset identifies the primary endpoint and its assessment framework but does not provide a numerical ORR estimate.
- Incomplete registry endpoint wording: the registry-reported time frame ends with “until,” and the registry-reported partial-response definition ends after “greater than or equal to.” No additional wording is inferred.
- Analysis population is not numerically specified: enrollment is 667, but the ClinicalTrials.gov record does not state the number of participants included in the formal response analysis.
- Safety data are cohort-specific: serious-adverse-event counts are reported by treatment-line cohort, not as randomized treatment comparisons.
- No comparative effect estimate: because no statistical-analysis records are reported, this page does not report an ORR confidence interval, p-value, odds ratio, risk ratio, or other formal treatment-effect estimate.
- Response assessment is not equivalent to survival: ORR measures tumor response according to the registered framework and does not by itself establish an effect on overall survival or another time-to-event endpoint.
13. Why This Trial Matters Statistically
BIRCH is a useful teaching example because it shows how the statistical interpretation of a clinical trial is shaped by its design architecture before any numerical result is considered.
| Concept | How it appears in BIRCH |
|---|---|
| Single-group design | The registry identifies a SINGLE_GROUP design with one arm. |
| Binary endpoint | The registered primary endpoint is objective response rate. |
| RECIST assessment | ORR is assessed according to RECIST v1.1. |
| Independent review | The primary endpoint specifies assessment by an Independent Review Facility. |
| Repeated assessment | Response is evaluated on a scheduled basis over time. |
| Confirmed response | The endpoint is based on confirmed best overall response. |
| Binomial estimation | A response percentage can naturally be represented as a proportion of responders among evaluable participants. |
| Confidence intervals | Precision around a response proportion would normally be communicated with a binomial confidence interval. |
| Comparative inference | A single-group study does not provide a randomized treatment-versus-control effect estimate. |
| Safety denominators | Serious adverse events are reported as affected participants over participants at risk within treatment-line cohorts. |
14. Interpreting an Objective Response Rate Correctly
An objective response rate is a proportion of participants meeting a prespecified tumor-response definition. It is not a continuous measure of average tumor shrinkage, and it is not a hazard ratio.
ORR does not mean that the same percentage of participants necessarily experienced a corresponding improvement in survival. It also does not describe how long a response lasted unless duration of response is analyzed separately.
The endpoint depends on RECIST v1.1 classification, independent review, repeated assessments, and confirmation of best overall response. The statistical percentage is therefore the final summary of a structured clinical-assessment process.
In a randomized trial, an ORR difference can directly compare response frequencies between assigned treatment groups. In a single-group study, an observed ORR has to be interpreted without that randomized concurrent comparison.
15. Statistical Inference Without a Randomized Comparator
There are several distinct questions that can be asked of a single-group response study, and they should not be conflated.
| Question | Statistical object | What it requires |
|---|---|---|
| What proportion responded? | Observed ORR | Responder count and defined analysis denominator |
| How precise is the ORR? | Confidence interval for a proportion | Observed response count and analysis denominator |
| Is ORR different from a reference value? | One-sample hypothesis test | A prespecified reference proportion |
| Is atezolizumab better than another treatment? | Comparative treatment effect | A valid comparator and an appropriate comparative design or analysis |
| Are treatment-line cohorts different? | Between-cohort comparison | Appropriate cohort definitions and a prespecified comparative analysis |
The registry-reported BIRCH data establish the first type of endpoint and identify the treatment-line cohorts, but they do not supply a formal statistical-analysis record for these comparisons. The distinction is important because a statistically precise estimate of one proportion is not equivalent to evidence of comparative superiority.
16. Results-Reporting Quality and Reproducibility
A strong clinical-trial statistical report should allow a reader to reconstruct what was analyzed, in whom, and under which statistical framework. For BIRCH, the ClinicalTrials.gov record provides several important pieces of that structure: the single-group design, enrollment, intervention, binary primary endpoint, response definition, assessment schedule, and serious-adverse-event counts.
Clearly specified
Study phase, enrollment, design model, intervention, primary endpoint type, RECIST v1.1 framework, independent review facility, and response-assessment schedule.
Not numerically reported
The numerical primary ORR result, confidence interval, p-value, and formal treatment-effect estimate.
Safety information reported
Serious-adverse-event counts and at-risk denominators for the three treatment-line cohorts.
Interpretive consequence
The page can explain the appropriate statistical framework without presenting unverified numerical efficacy results.
17. Trial History
Trial start
The registry lists January 22, 2014 as the study start date.
Primary completion
The registry lists May 28, 2015 as the primary completion date.
Final registry status
The trial is listed as completed, with results posted and 23 outcome measures reported in the ClinicalTrials.gov record.
18. What Can Be Concluded From the Supplied Data
The ClinicalTrials.gov record supports several precise conclusions about the statistical structure of BIRCH.
- The study is a phase 2, single-group clinical trial with one registered intervention, atezolizumab.
- The enrolled population consists of participants with PD-L1-positive locally advanced or metastatic non-small cell lung cancer.
- The registered primary endpoint is a binary objective-response measure based on confirmed best overall response of CR or PR according to RECIST v1.1 and assessed by an independent review facility.
- The response endpoint uses repeated assessments beginning at screening, followed by the registry-specified assessment schedule.
- The ClinicalTrials.gov record does not contain a formal statistical-analysis record for the primary endpoint.
- The ClinicalTrials.gov record therefore do not establish a numerical ORR estimate, confidence interval, p-value, or comparative treatment effect that can be reproduced here.
- Serious-adverse-event counts are available for the three reported treatment-line cohorts and should be interpreted descriptively.
19. Related Statistical Concepts
Learn more about the methods used in this trial:
20. Related Statistical Calculators
21. Sources
- ClinicalTrials.gov: BIRCH, NCT02031458.
- PubMed: PubMed record for PMID 38367405.
- PubMed: PubMed record for PMID 33241650.
Continue through the Clinical Biostats statistical library
Explore the underlying statistical concepts through tutorials and use appropriate calculators for binary endpoints, confidence intervals, and clinical-trial analysis.
22. Record Summary
BIRCH is a phase 2, single-group study of atezolizumab in participants with PD-L1-positive locally advanced or metastatic non-small cell lung cancer. Its registered primary endpoint is objective response rate, defined using confirmed best overall response of partial response or complete response according to RECIST v1.1 and assessed by an independent review facility. The endpoint is binary, making the natural statistical framework a binomial response analysis with a response proportion and an appropriate confidence interval. However, the ClinicalTrials.gov record contains no formal statistical-analysis records and no numerical ORR estimate, confidence interval, or p-value. The available serious-adverse-event information is cohort-specific and descriptive. The central statistical lesson is therefore not a particular treatment-effect estimate, but the importance of matching the inferential method to the study design, endpoint type, analysis population, and level of evidence actually reported.